Interfibrillar stiffening of echinoderm mutablecollagenous tissue demonstrated at the nanoscaleJingyi Moa, Sylvain F. Prévostb, Liisa M. Blowesc, Michaela Egertovác, Nicholas J. Terrilld, Wen Wanga,e,Maurice R. Elphickc,1, and Himadri S. Guptaa,e,1

aSchool of Engineering and Material Science, Queen Mary University of London, London, E1 4NS, United Kingdom; bBeamline ID02, European SynchrotronRadiation Facility, Grenoble 38000, France; cSchool of Biological and Chemical Sciences, Queen Mary University of London, London, E1 4NS, UnitedKingdom; dBeamline I22, Diamond Light Source, Harwell Science and Innovation Campus, Harwell, OX11 0DE, United Kingdom; and eInstitute ofBioengineering, Queen Mary University of London, London, E1 4NS, United Kingdom

Edited by David A. Weitz, Harvard University, Cambridge, MA, and approved August 29, 2016 (received for review June 12, 2016)

The mutable collagenous tissue (MCT) of echinoderms (e.g., seacucumbers and starfish) is a remarkable example of a biologicalmaterial that has the unique attribute, among collagenous tissues,of being able to rapidly change its stiffness and extensibility underneural control. However, the mechanisms of MCT have not beencharacterized at the nanoscale. Using synchrotron small-angle X-raydiffraction to probe time-dependent changes in fibrillar structureduring in situ tensile testing of sea cucumber dermis, we investigatethe ultrastructural mechanics of MCT by measuring fibril strain atdifferent chemically induced mechanical states. By measuring avariable interfibrillar stiffness (EIF), the mechanism of mutability atthe nanoscale can be demonstrated directly. A model of stiffnessmodulation via enhanced fibrillar recruitment is developed to explainthe biophysical mechanisms of MCT. Understanding the mechanismsof MCT quantitatively may have applications in development of newtypes of mechanically tunable biomaterials.

mutable collagenous tissue | synchrotron small-angle X-ray diffraction |nanoscale mechanics | fibrillar deformation | sea cucumbers

The mechanical properties of biological tissues are usuallyoptimized to operate within specific physiological loading

and strain ranges (1, 2). With the exception of the commonphenomenon of strain stiffening that occurs during mechanicalloading (3), material-level changes in the overall mechanicalproperties of tissues typically occur slowly, driven by growth,remodeling, or aging (4). The molecular-level mechanisms under-pinning these changes often involve permanent, irreversible changes,including covalent cross-linking via disulphide bridges in tendon (5),formation of metal-ion/protein complexes (6), or replacement ofwater with an inorganic phase as in biomineralization (7), althoughviscoelastic mechanical responses may involve transient cross-linking(8). In contrast, changes in the mechanical properties of animaltissues that occur actively and reversibly within a few seconds arecanonically mediated by ATP-dependent molecular motors, as inmuscle (9). A notable exception is the mutable collagenous tissue(MCT) of echinoderms (e.g., starfish, sea urchins, sea cucumbers),which undergoes rapid changes in stiffness under the control of thenervous system via ATP-independent mechanisms (10–12). MCT isubiquitous in echinoderms (12), for example, in the dermis (skin) ofsea cucumbers (13, 14), in the compass depressor ligament (CDL) ofsea urchins (15–17), and in the arms of feather stars (18). Thepresence of MCT enables functionally diverse behaviors; for exam-ple in starfish, MCT enables body wall stiffening during feeding onprey and it also enables irreversible body wall softening before armautotomy as a defense against predation (12). Thus, MCT representsan evolutionary adaptation of collagenous tissue to change me-chanical properties dynamically, whereas in other phyla collagenoustissues largely act as passive mechanical springs. The benefits ofMCT also include a much lower energy expenditure (19) comparedwith muscle tissue and the presence of MCT is considered to havebeen a major factor in the evolutionary success and ecological di-versity of echinoderms [reviewed in Barbaglio et al. (20)].

The initial identification of connective tissue of echinodermsas having mechanically unusual properties—illustrated by itsdenotation as “catch” connective tissue—was through the observedstiffening and softening response of such tissues to sea water ofdifferent ionic compositions, as well as neurotransmitters (e.g.,acetylcholine) (14) and drugs (e.g., cocaine) (21). Such chemicalmeans to induce mutability in MCT remain a convenient and re-producible method to induce mechanically altered states (15, 22,23). Specifically, previous studies demonstrated that alteration ofextracellular Ca2+ and K+ levels modulates the stiffness of livingtissues in sea urchin spine ligaments (14, 24–28), holothuriandermis (13, 22, 29, 30), and starfish (30). Increased K+ concen-tration increases the stiffness of these particular examples of MCT,whereas decreased Ca2+ concentration lowers stiffness, comparedwith artificial sea water (ASW) as a reference solution. Thechanges induced by these chemical methods are within the sameorder of magnitude as physiologically relevant mechanical changesoccurring in MCT in vivo: Under mechanical stimulation(pressing the tissue by hand), starfish body wall MCT stiffens bya factor of ∼3.3 (30), whereas increasing K+ concentration leadsto an increase of ∼6.1 (30). Conversely, the reduction of stiffnessin calcium-free artificial sea water is considerable (31), which iscomparable to the reduction to zero in the extreme case of limbautotomy where the tissue disintegrates structurally (32).

Significance

Collagen plays crucial biomechanical roles in a wide array of ani-mal tissues, but its mechanical properties remain largely staticover short timescales. However, echinoderms (sea cucumbers,starfish) are striking exceptions to this rule, having “mutablecollagenous tissue” with changeable mechanical properties, en-abling complex locomotion, postural maintenance, defense, andreproductive strategies. Using a high-resolution X-ray probe thatmeasures how the building blocks—fibrils—of echinoderm con-nective tissue stretch, slide, or reorient in real time, we show thatsea cucumbers achieve this remarkable property by changing thestiffness of the matrix between individual fibrils, rather than theproperties of the fibrils themselves. Understanding the mecha-nisms of mutability in this unique tissue may help design novelmechanically tunable synthetic biomaterials.

Author contributions: J.M., M.R.E., and H.S.G. designed research; J.M. and H.S.G. per-formed research; S.F.P., L.M.B., M.E., N.J.T., and H.S.G. contributed new reagents/analytictools; J.M. and H.S.G. analyzed data; J.M., S.F.P., L.M.B., M.E., N.J.T., W.W., M.R.E., andH.S.G. wrote the paper; and W.W. and H.S.G. supervised research.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. Email: h.gupta@qmul.ac.uk or M.R.Elphick@qmul.ac.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1609341113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1609341113 PNAS Early Edition | 1 of 10

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

ENGINEE

RING

PNASPL

US

The unusual mechanical properties of MCT must arise from themicro- and ultrastructure of this tissue, which shows both com-monalities with, as well as some clear differences from, the morefamiliar vertebrate collagenous tissues such as skin, tendon, andbone. At the molecular level, the collagen of sea cucumber MCT isdifferent from heterotrimeric vertebrate type I fibrillar collagen,consisting of homotrimers with three α1 polypeptide chains (12).These collagen molecules aggregate into discontinuous spindle-shaped collagen fibrils with a mean diameter of ∼17 nm (33). Pro-teoglycans are bound to the fibrillar surfaces, which along withnoncollagenous proteins such as tensilin, stiparin, softenin, andfibrosurfin compose the interfibrillar matrix (34–36). Together withfibrillin-rich microfibrils (34) the fibrillar collagen network comprisesthe bulk of the extracellular matrix (ECM) of MCT. Dispersed inthis ECM are clusters of juxtaligamental cells (JLCs) (12, 15, 19, 24,28, 29) (as seen from transmission electron and light microscopy),which are MCT-effector cells that are under neural control (12, 15).Furthermore, it is the innervation of MCT that is a key distinctionbetween echinoderm and vertebrate collagenous tissues (15).Initial hypotheses about the ultrastructural mechanism enabling

the mechanical mutability described above focused on ion-mediatedcreation of physical cross-links between fibrils and within the

interfibrillar matrix in the ECM (11, 23). Divalent calcium ions wereproposed to be especially effective in increasing interfibrillar matrixstiffness, and their depletion in Ca2+-free sea water solutions wasbelieved to be a major reason for the reduced stiffness. However,this hypothesis was disproved when cell-lysed MCT showed nomechanical mutability in the presence of such solutions (35). It istherefore believed that the ionic treatments directly affect cellularsecretion pathways in the JLCs, inducing release of proteins thatalter interfibrillar binding, thus changing the stiffness of the tissue(12). Several such proteins, including tensilin, softenin, and novelstiffening factor (NSF), have been identified (31, 34, 35, 37, 38). It isbelieved that the JLCs secrete such effector proteins as a result ofexternal stimuli (such as touch, aggressive attack or alteration inthe ionic strength of the sea water around the animal), therebychanging the stiffness of the tissue. Biochemical evidence to sup-port this hypothesis includes the structural similarity of some ofthese proteins to the tissue inhibitors of matrix metalloproteinases(TIMPs) found in vertebrates (39). Consistent with this notion, ithas also been suggested that cysteine-rich sea urchin fibrillar do-mains (SURFs), found so far in the sea urchin collagen 2α and 5αN-propeptides, as well as fibrosurfin (an interfibrillar protein) (36,40), play a role in enabling mutability. The 2α N-propeptides andfibrosurfin colocalize on collagen fibril surfaces in adult sea ur-chins. However, despite this considerable level of biochemicalinsight into MCT (12), the biophysical mechanisms by which thealteration in mechanics is mediated by the nanostructure—whether at the fibrillar or the intrafibrillar level—are still notcompletely understood.Techniques used to correlate ultrastructure with mechanics have

been by necessity largely static and indirect, including imagingtissue after alteration of mechanical state with techniques liketransmission electron microscopy (15). These introduce unavoid-able artifacts from sample preparation and do not measure thechanges as they occur in real time. The use of in situ synchrotronX-ray diffraction to provide molecular- and supramolecular-levelimages of the ultrastructural conformation during alteration of themechanical state of MCT is a direct way to overcome these limi-tations (41–44). The axial periodicity of electron density along thelong axis of collagen fibrils (45), with a repeat distance of D ∼65–67 nm, leads to Bragg diffraction peaks in X-ray scattering in thesmall-wavevector domain (<5 nm−1) characteristic of small-angleX-ray diffraction (SAXD) (45). Shifts in these peaks, as would beinduced by mechanical loading or ionic treatments, are therefore ameasure of the nanoscale fibril strain demonstrated for vertebratetissues (41–43, 46, 47). By combining micromechanics with in situsmall-angle X-ray scattering, it has been possible to shed light onthe fundamental ultrastructural mechanisms enabling viscoelastic-ity, toughness, and force generation in vertebrate tissues rangingfrom tendon (48), bone (44, 46), and aorta (49) to muscle (50), aswell as more unusual examples of biological optimization such asarmored fish scales (43). Using SAXD, it was found that in cross-link–deficient fibrils, increased molecular slippage led to larger fi-bril strains, compared with normal collagen fibrils (48); that hightoughness of antler bone was due to inorganic/organic friction atthe intrafibrillar level (41); and that fibrillar reorientation bluntedcrack propagation in skin (51), among other examples.When combined with high-intensity synchrotron X-ray sources,

time-resolved SAXD with in situ micromechanical loading could beused to quantify the fibrillar deformation mechanisms of MCT invarious states of mechanical mutability, thus clarifying the biophysicalmechanisms enabling this remarkable behavior. Here, we apply thesetechniques to the sea cucumber dermis as a model system (Fig. 1).Understanding the molecular mechanisms enabling mutability mayhave applications in developing dynamic biomaterials, systems ca-pable of changing their mechanical properties, and the design ofmechanically tunable implants. The adaptive mechanical propertiesin MCT could, for example, provide insight into the repair of con-nective tissue pathologies in soft tissue, such as therapy in tendon or

Fig. 1. Sea cucumber body wall MCT. (A) Sea cucumber H. leucospilota.(B) Transverse section of sea cucumber body wall stained using Masson’s tri-chromemethod. Collagen fibrils appear blue. (C) The sheet of bodywall after theanimal was cut in half along the longitudinal plane. The blue rectangle indicatesthe dimensions and location of the sectioned specimen, with the long dimensionalong the longitudinal axis. (D, Upper) View of sectioned sea cucumber dermisincluding dark outer dermis and inner layer. (D, Lower) The tensile test specimen.Before testing, the dark-pigmented outer dermis as well as the inner layer wasremoved, leaving only the center part of the specimen. (E) Time-dependentchange in sea cucumber MCT mechanics induced via ionic treatment. The peakstress (per cycle) is plotted during strain-controlled cyclic loading of sea cucumberdermis at 0.3 Hz (to 15% tissue strain), with tissue immersed in ASW until ∼290 s,followed by a change of the immersing solution to KASW (stiffening agent). Aclear rise of peak stress (per cycle) is observed, fitted with a sigmoidal curve as aguide to the eye. Insets (E, Right) show a magnified time range over a few(seven) cycles, with both maximum and minimum stress and strain indicated.

2 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1609341113 Mo et al.

ligament weakening resulting from surgery or immobilization (12);into the design of implants capable of generating active forces; andinto the area of neural implants where variable stiffness during in-sertion and implantation has been proposed (52).

ResultsSynchrotron SAXD measurements of fibrillar strain in MCTfrom sea cucumber dermal tissue were carried out at beamline

ID02 at the European Synchrotron Radiation Facility as shown inFig. 2. Before testing, tissue specimens were chemically incubatedin ASW and two ionically modified solutions of ASW [potassium-rich ASW (KASW) and calcium-free ASW (CaF-ASW)], whichare known to induce standard state, stiffening and softening of seacucumber dermis, respectively (22, 23) (details in Materials andMethods). The fibril strain «F is the fractional increase in fibrillength (measured from the shifts in the meridional Bragg peaks in

Fig. 2. In situ nanomechanics with synchrotron SAXD. (A) Experimental configuration: Tensile tester (Center) with MCT specimen mounted along the X-ray beampath in transmission geometry with CCD detector (Left). (A, Right, Inset) Magnified view of sample in chamber and incident X-ray beam (right) with SAXD scatteringshown on left. The tensile strain is applied along the vertical direction. (A, Upper Right) Schematic of body wall of sea cucumber shown in Fig. 1, with tensile testspecimen sectioned with long axis parallel to the long axis of the animal. (B) Data reduction pipeline: (B, 1) A 2D SAXD pattern from collagen fibrils in sea cucumberdermis MCT with predominant fibril orientation vertical. Radial (q) direction is indicated. Dotted lines denote the ring over which the azimuthal averaging of intensityis carried out. (B, 2) The azimuthally averaged radial intensity profile I(q) for the pattern in B, 1. (C, 1) The same 2D SAXD pattern as in B, 1, with the inner, outer, andcentral rings (i, o, and c, respectively) shown schematically, over which radial averaging of intensity is carried out. Azimuthal (χ) direction is indicated. (C, 2) The radiallyaveraged intensity profile I(χ). In B, 2 and C, 2 both experimental data (solid circles) and fits to model functions (solid lines) are shown. (D) Radial intensity profile I(q)for three levels of applied tissue strain «T = 0% (circles), 30% (squares), and 60% (diamonds), showing the shift of peak position to lower wavevector with increasingstrain. (E) Tissue stress–tissue strain plot for sea cucumber dermis in tension, with circles (black, 0%; red, 30%; blue, 60%) indicating the points fromwhich the I(q) plotsin D are shown. (F) Flowchart corresponding to the data reduction steps in B, 1–C, 2, with parameters obtained at each step indicated.

Mo et al. PNAS Early Edition | 3 of 10

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

ENGINEE

RING

PNASPL

US

the SAXD pattern of MCT collagen fibrils), whereas tissue strain«T is the fractional increase in MCT sample length and tissue stressσT is the force divided by sample area (Materials and Methods).Considering the fibrillar-level strain «F and tissue stress developedin MCT during stretch to failure tests, it is observed (Fig. 3A) thatfibrils in tissues with different chemical stimulation—CaF-ASW,ASW, and KASW—show a differing extent of elongation at thesame tissue strain «T. At a given tissue strain, the amount of fibril

strain is proportional to the stress taken up by the fibrils. At a tissuestrain «T around 10%, fibrils in KASW tissue have a much higherextension of ∼0.5% compared with ASW (0.07%). Likewise, thefibril strain for CaF-ASW is much lower (∼0.001%), indicatingmainly interfibrillar sliding. The maximum fibril strain developed inKASW is much larger than in CaF-ASW, and the fibril strain inASW is in between that of KASW and CaF-ASW. Fig. 3B showscorresponding averaged mean strain–stress curves for body walltissue in KASW (red), ASW (black), and CaF-ASW (blue). Theaveraged macroscopic stress σT with tissue strain «T at 40% forKASW-treated (4.08 MPa) specimens is significantly higher thanfor CaF-ASW (0.13 MPa), whereas the ASW-treated (0.77 MPa)specimens are in between these extremes. We note that the in-crease of fibril strain with applied tissue strain is not completelysmooth in all cases, as evidenced by the error bars. This is mostnoticeable in the case of KASW-treated tissue, where there areclear local peaks at ∼10% and 30% strain followed by dips. Thestructural reasons for this behavior are considered further when themodel to explain fibrillar deformation is developed in Discussion.Further, the variation of the macroscopic tensile stress (at the tis-sue level) for the whole specimen for all treatments (KASW, ASW,and CaF-ASW) is much greater compared with the strain de-veloped at the fibril level (Fig. 3A), indicating that interfibrillarcomponents of the extracellular matrix are important in mecha-nisms of MCT.The differences in Ca2+ or K+ concentrations in CaF-ASW,

ASW, and KASW led to changes of maximum tangent modulusand maximum tissue stress (Fig. 4). Ca-FASW–treated (0.69 ±0.59 MPa) samples had ∼80% lower maximum tangent moduluscompared with ASW (3.23 ± 0.40 MPa), whereas the maximumtangent modulus for KASW-treated samples (17.27 ±6.70 MPa)was four times larger. Concurrently, the maximum stress of eachstate is very different, with the loads borne by CaF-ASW–treatedtissue (0.30 ± 0.32 MPa) being much less (∼80%) compared withASW tissue (1.30 ± 0.20 MPa) whereas KASW tissue (6.39 ± 0.44MPa) had maximum stress four times higher than the control.Similarly, nanoscale parameters like fibril strain also show cleardifferences between treatments (Fig. 4 C and D). Compared withmaximum fibril extension (Fig. 4C) developed in KASW (0.94%),fibrils in CaF-ASW (0.09%) and ASW (0.35%) had a reducedelongation, with ∼95% and 80% less strain, respectively. Fig. 4Dshows the ratio of fibril strain to tissue strain («F/«T) for the threestates, which is observed to be consistent with Fig. 4C, showing thatstiffened tissue sections have («F/«T) of 0.044, larger than controlspecimens, whereas softened tissue sections have almost negligiblefibril strain take-up («F/«T ∼ 0). The parameter «F/«T is used toconfirm the modeling results, which are illustrated in Model andDiscussion.In a complementary manner, we consider the alterations in the

fibril orientation distribution on application of external load.The υ parameter, derived from the angular distribution of theSAXD intensity, is a dimensionless number that is zero whenfibrils are distributed at all angles with equal likelihood andpositive when fibrils are aligned either along one or along acouple of principal directions (Materials and Methods). A rep-resentative plot of the υ parameter as a function of tissue strain(Fig. 5A) exhibits an initially high value of ∼1.3 [correspondingto two main fibril directions equidistant (azimuthally) from thevertical direction] that is followed by a decrease to a minimumaround 20% tissue strain. The initial two directions correspondto two principal helical fiber pitches along the long axis of theanimal. The initial reduction in the υ parameter corresponds to amore random fibril orientation, indicated schematically in Fig. 5A–C,and represents the stress-induced breakdown of the two main fibrildirections into a single broad distribution centered on the (vertical)direction of applied tensile load. For tissue strains larger than 20%,the υ parameter increases monotonically with increasing tissuestrain, before leveling off near tissue strains of ∼40–50% close to

Fig. 3. Altered fibrillar stress and strain take-up in ionically treated MCT.(A) Fibril strain vs. applied tissue strain from ionically treated sections ofMCT dermis, measured from the peak shifts of the fifth-order collagenreflections in the SAXD pattern. The rate of increase of fibril strain withtissue strain («F/«T) is proportional to the amount of stress taken up by thecollagen fibrils. Data are from control (ASW) (black circles, n = 4), stiffened(KASW) (red squares, n = 4), and softened (CaF-ASW) (blue diamonds, n = 3)MCT. All samples in each group are binned according to tissue strainwith bin widths of 2.0%; error bars are SDs. Stiffened MCT exhibits ahigher rate of increase of fibril strain compared with control, whereassoftened MCT shows essentially no increase in fibril strain. Inset sche-matics: (A, i ) Fibrils (striated ellipsoids; length LF0) separated by interfi-brillar matrix, in unloaded MCT of length LT0; (A, ii) in softened MCT(CaF-ASW treated), while the tissue elongates («T > 0), the fibrils do notstretch, but slide in the interfibrillar matrix («F = 0); (A, iii) in stiffened MCT(KASW treated), there is increased stress transfer to the fibrils, leading tofibrillar stretching («F > 0). (B) Corresponding macroscopic tensile stress/strain curves for the control, stiffened, and softened groups, binnedaccording to tissue strain (error bars: SDs), showing clear differences intangent modulus and maximum stress achieved.

4 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1609341113 Mo et al.

macroscopic failure. The increase of the υ parameter represents anarrowing of the azimuthal width of the initial broad fibril distri-bution around the direction of applied load. The tissue strain at thetransition between the reduction of the υ parameter and the sub-sequent increase is denoted by «Tr. The analysis of mean «Tr acrossthree groups of specimens is shown in Fig. 5D. The results indicatethat soft (CaF-ASW) samples always have a higher «Tr, relative tothe control (ASW) and stiffened (KASW) samples, which impliesthat the rate of reorientation (from the initial distribution with twomain directions to a random and then highly oriented distribution)is slower for the softened MCT.

Model and DiscussionThese experimental results, showing clear alterations in the de-formation at the fibrillar level when MCT is stimulated into its stiffand soft states, can be used to build a simple model that sheds lighton the key biophysical mechanisms enabling mutability (31, 34, 35,37, 38). Prior research has proposed, but not directly demon-strated, that certain proteins secreted by JLCs including tensilin(35, 37), stiparin (34), and NSF (31) act to cross-link the fibrils. Acomplete ultrastructural mechanism for MCT mutability has not,however, been quantitatively established, and each protein appearsto be involved only in a specific subset of the mechanical response,such as in stiffening a compliant specimen back to the referencestate [as is the case for tensilin (35, 37)]. At the ultrastructural level,the parallel-packed fibrils and interfibrillar matrix of MCT can berepresented as shown in Fig. 6A. The covalently cross-linked (α1)3collagen fibrils are expected to have much greater stiffness [∼0.5–2 GPa (53)] than the interfibrillar matrix [which can be considereda negatively charged hydrated gel (54)], although precise values forthe interfibrillar matrix modulus are unknown. Under tensileloading, the highly anisotropic fibrillar structure together with theexpected stiffness mismatch between the fibrils and the interfi-brillar matrix will lead to a characteristic inhomogeneous de-formation field at the nanoscale. In this deformation field, tensileforces develop in the fibrils and matrix, and significant shear occurs inthe interfibrillar matrix connecting fibrils (55). The shearing forcelines are shown in Fig. 6B, Insets, and can be considered as repre-

sentations of the cross-linking between fibrils proposed previously(12). Consequently, the fibril strain is only a fraction of the total strain(due to the remaining shearing strain in the interfibrillar matrix).To keep the model analytically simple we consider a unidirec-

tional fiber composite; whereas the initial unstrained MCT showstwo main fibril directions around the direction of stretch (Fig. 5B),it is observed that for tissue strains larger than ∼20% the fibrils arehighly aligned to the direction of applied stress and the uniaxialfibril arrangement is expected to be a good approximation for thisregion at least. This type of model, denoted a staggered model,has been proposed before, by us (41) and others (55, 56) for de-formation of the ultrastructure of bone mineralized fibrils, tendon,enamel, and dentine (46, 57), where a similar high-stiffness elementin tension (e.g., mineral platelet) is effectively in serial loading witha low-stiffness element loaded in shear (e.g., collagen fibrils). Thedeformation of the fibril «F, shear of the interfibrillar matrix γIF, andtissue level stress σT, among other quantities, have been calculatedfrom load-balance equations at the nanoscale (41). These lead toexpressions for the fibril to tissue strain ratio, and the tissue mod-ulus ET, in terms of the structural and constitutive parameters

«T«F

= 1+4ρ21

1−Φ1

Φ1

EF

γIFEIF, [1]

ET =Φ1EF«F«T

+ ð1−Φ1ÞEIF =Φ1EF

1+ 4ρ21

1−Φ1Φ1

EFγIFEIF

+ ð1−Φ1ÞEIF .

[2]

In the equations above, Φ1 denotes the fibril volume fraction, ρ1the fibril aspect ratio, EF the fibril elastic modulus, and EIF theinterfibrillar modulus, and γIF ∼ 0.40 is the ratio between shear(GIF) and tensile (EIF) modulus of the interfibrillar matrix (GIF =γIFEIF) (41); other terms have been defined earlier in the text. Asour scheme enables measurement of deformation at the fibrillarlevel concurrently with tissue-level mechanical stress and strain,a parametric variation of fibril strain and tissue modulus may

Fig. 4. Quantified MCT mechanics. (A and B) Averaged maximum tangent modulus (A) and maximum stress (B) for control (ASW; black; n = 4), stiffened(KASW; red; n = 4), and softened (CaF-ASW; blue; n = 3) specimens from MCT dermis. Error bars are SDs. n denotes the number of samples in each treatmentgroup. At the tissue level, stiffened MCT always exhibits remarkably higher mechanical properties compared with the control group, and for the softenedone, the binned maximum tangent modulus and the maximum stress are almost negligible. (C and D) At the fibrillar level, the fibril strain (C) and the ratio ofthe fibril strain to tissue strain (D, the fraction of the deformation taken up at the fibril level) for the stiffened, control, and softened MCT follow the sametrend with properties at the tissue level.

Mo et al. PNAS Early Edition | 5 of 10

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

ENGINEE

RING

PNASPL

US

now be carried out and compared with the experimental resultsreported earlier in Figs. 4 and 5. These results are shown in Fig.6B and discussed below.Considering, for the moment, that the collagen fibrils have a

constant elastic modulus EF, we first examine the effects of al-tering the interfibrillar modulus on the change in tissue stiffness.Such a scenario corresponds to an alteration in cohesion due tosecretion of stiffening factors such as tensilin, stiparin, and NSF.As seen in Eq. 2, the increase in tissue stiffness arises due to boththe increased load borne by the interfibrillar matrix (second termon the right-hand side in Eq. 2) and—much more significantly—the increased stress borne by the collagen fibril due to the largershear stress transferred by the interfibrillar matrix (first term on the

right-hand side). The stress borne by the collagen fibril is large dueto the large contact surface area between the fibrils and theinterfibrillar matrix due, in turn, to the large fibril aspect ratio ρ1.Concurrently with the increase in tissue stiffness, the fibril strainincreases as a fraction of the tissue strain, as larger tensile forcesare transferred to the elastic fibrils with increased fibril strain.To compare the model predictions with experimental data

(Fig. 6 B and C), initial estimates of some of the unknownparameters need to be made. The length and fibril diameterdistributions for echinoderm collagen fibrils have been estimatedpreviously (33, 58). Whereas the maximum and minimum valuesspan a wide range, a constant spindle-shaped morphology wasreported (33). The most frequent value (mode) in the diameter

Fig. 5. Strain-induced fibril alignment of MCT in stiffened, control, and softened states: (A–C, Upper) Schematic illustration of fibril distribution at increasingtissue strain levels. (A–C, Lower) Corresponding I(χ) plots for a control (ASW) sample with corresponding υ parameter. The vertical arrow to the left of theplots indicates the direction of applied tissue strain. A (0% strain) shows bimodal distribution due to two groups of fibers (I and II) inclined to the direction oftensile load at two principal fiber directions (arrows). In the I(χ) profile (A, Lower), the two sets of peaks arising from these two groups of fibers are labeled Iand II, respectively. B (20% strain) shows a wider range of orientations due to fibrils progressively reorienting (indicated by the red arcing arrows) toward thetensile axis (low υ) and C (55% strain) shows highly aligned fibers along the vertical direction (higher υ). (D) Initial decrease followed by increase of the υparameter with increased tissue strain, for control (ASW) MCT, exhibiting an initial decrease and a local minimum at tissue strain ∼20%, followed by anincrease. (E) A typical stress–strain curve for ASW-treated MCT. In D and E, rectangles indicate strain locations corresponding to A–C. (F) Variation of thestrain-induced changes in the υ parameter as a function of the mechanical state of MCT due to ionic treatment. Data from one representative MCT specimenin each state are shown. The tissue strain corresponding to the local minimum in υ is indicated by a vertical arrow, denoted «Tr, and is lowest for the stiffenedand highest for the softened specimen. (G) Averaged «Tr across the three treatments [control (ASW; black; n = 4), stiffened (KASW; red; n = 4), and softened(CaF-ASW; blue; n = 3)]. Error bars are SDs.

6 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1609341113 Mo et al.

distribution was ∼100 nm (58), and the length ranged from50 μm to 1 mm. Using the constant shape of the collagen fibrilsreported previously (33) and the diameter distribution shown inref. 58, these values can be used to estimate the maximal lengthto be ∼120 μm. These lead to an aspect ratio of ∼1,000. The fibrilmodulus for individual sea cucumber fibrils isolated from thetissue was measured to be ∼500 MPa (53). As described inSupporting Information, these values can be used, together withthe experimentally determined values of «F/«T and ET, to obtainestimates of fibril volume fraction Φ1 ∼ 0.54 and the interfibrillarmodulus EIF ∼ 50 × 10−6 MPa (in the ASW case). The fibrilvolume fraction will not vary across treatments, but the interfi-brillar modulus will change. With these numerical values, the twotrends described above—increase in tissue stiffness (ET) as wellas in fibril strain ratio «F/«T with increasing EIF—are plotted inFig. 6B, together with the experimentally measured values forthe softened, control, and stiffened groups. Eq. 1 is used tocalculate interfibrillar matrix stiffness EIF for each tissue groupfrom measured «F/«T, which are then plotted (symbols) togetherwith model curves (lines). In a similar manner, Eq. 2 is used toobtain EIF from measured tissue moduli ET for each group. Forconsistency of the model, the two calculated sets of EIF shouldmatch, and indeed it is observed that both methods give similarvalues for EIF across the tissue groups.In principle, however, there are two distinct ways in which

the stiffness of MCT can be modified: either by the interfibrillarmatrix stiffening (considered above) or by alterations of themechanics of the collagen fibrils, possibly by modulation ofintrafibrillar cohesion. These two scenarios—fibrillar vs. inter-fibrillar stiffening—lead to different behaviors at the fibrillarlevel. In the fibrillar stiffening case, the fibrillar strain (as afraction of tissue strain) will reduce as the tissue stiffens,whereas in the interfibrillar stiffening case, the fibrillar strainwill increase as the tissue stiffens, as can be seen by combiningEqs. 1 and 2 to obtain parametric plots of «F/«T and ET asfunctions of EIF and EF

«F«T

=ET − ð1−Φ1ÞEIF

Φ1EF. [3]

In the model, changing the aspect ratio by ±10% leads to avariation of 20% in both «F/«T and ET (Supporting Information,Sensitivity of Fibril Strain Ratio and Tissue Modulus to ModelParameters). To conclusively demonstrate interfibrillar stiffeningand exclude fibrillar stiffening as the key mechanism for mechan-ical changes in MCT, the experimental data for «F/«T and ET areplotted in Fig. 6C together with two sets of three predictivecurves from the model. The first set of three curves (solid lines)corresponds to a continuous increase in interfibrillar modulus(for several discrete values of fibrillar stiffness), whereas thesecond set (dashed lines) corresponds to continuously increasingfibrillar stiffness for several values of interfibrillar stiffness. It isclearly seen that the experimental data show an increase in «F/«Twith ET, corresponding to the case of interfibrillar matrix stiff-ening. Further, it is apparent that the mean values lie along thepredicted curve for a collagen fibril modulus EF = 600 MPa,which is close to the value of 500 MPa reported by Eppellet al. (53). This finding provides evidence in support of thelong-held, but not directly demonstrated, hypothesis that thealteration of MCT mechanical properties arises due to changesin interactions between fibrils (through changes in the interfibrillarmatrix) rather than alterations in the mechanical properties ofthe fibrils themselves (12).From the variation of interfibrillar matrix stiffness certain

observations can be made. In MCT, EIF is, even in the stiffeststate, ∼0.25 kPa—at least six orders of magnitude lower than thestiffness of the fibril ∼0.6 GPa. In the state of least stiffness,

Fig. 6. Staggered model of MCT nanomechanics. (A) Staggered model forMCT: discontinuous, spindle-shaped collagen fibrils aggregating in parallel.The attached proteoglycans serve as a binding site for interfibrillar cohesionmediated by cross-linker molecules (12). (B) Elevation of interfibrillar stiff-ness EIF leads to a corresponding increase in both tissue modulus ET (left-hand abscissa; solid line, staggered model prediction) and fibril-to-tissuestrain ratio («F/«T) (right-hand abscissa; dashed line, staggered model pre-diction). Symbols show experimental values for each tissue group with EIFcalculated from the staggered model equations (solid symbols from Eq. 1;open symbols from Eq. 2). Open arrows indicate the abscissa each line be-longs to. B, Inset schematics show the shear transfer, and consequent stresstake-up, between fibrils in softened (Left) and stiffened (Right) states. In-creased shear stress in the stiffened interfibrillar matrix is shown qualita-tively by a larger number of interfibrillar shear lines. (C) Positive correlationbetween «F/«T and ET demonstrates that interfibrillar stiffening is themechanism for alteration of MCT mechanics. The staggered-model relationsbetween («F/«T) and ET (Eq. 3) are shown via solid curves (with a positivegradient) for varying EIF and three fixed levels of EF (blue, 60 MPa; black,600 MPa; and gray, 6,000 MPa). Likewise, dashed lines show staggered-model predictions for («F/«T) vs. ET for varying EF and three fixed levels of EIF(black, 50 × 10−6 MPa; blue, 250 × 10−6 MPa; and gray, 500 × 10−6 MPa).Symbols show experimental («F/«T) and ET, which show a clear positive cor-relation, indicating interfibrillar stiffening (blue, CaF-ASW; black, ASW; andred, KASW; error bars are SDs).

Mo et al. PNAS Early Edition | 7 of 10

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

ENGINEE

RING

PNASPL

US

there is a factor of ∼100 reduction in EIF relative to the alreadylow value of 0.25 kPa, to ∼5–10 Pa, and the matrix can be con-sidered a fluid. However, the increased stiffness of MCT in thestiffened state [necessary for its physiological maintenance ofposture or locomotion (12)] is almost entirely due to the in-creased fibrillar recruitment to bear stress and not due to thestress carried in the interfibrillar matrix. The ratio of the stress inthe interfibrillar matrix to that in the fibril can be calculated from

σIFσF

≈�1−Φ1

Φ1

��EIF

EF

��«T«F

�=�0.460.54

��50× 10−6

0.6

��1

0.01

�

= 7.1× 10−3;

i.e., only ∼0.7% of the total stress is borne by the interfibrillarmatrix. Under alteration of mechanical state, the relative in-crease in fibril stress can be calculated from the change in fibrilstrain ratio, via

σF =Φ1EF«F«T

.

We observe that fibril stress increases by a factor of ∼4 (Fig. 4D) inthe stiff state compared with the standard state. Conversely, in thesoftened state, the fibril stress decreased by a factor of ∼10 relativeto the standard state. We also note that we have (for simplicity)considered the interfibrillar matrix to stiffen homogeneously. Inpractice, there can be local heterogeneities (both temporal andspatial) in stiffness, possibly due to conformational changes inthe noncollagenous proteins, local entanglements, and other phe-nomena. Such heterogeneities would lead to local increases (ordecreases) in interfibrillar stiffness, which would increase (or de-crease) stress transfer between fibrils. Such alterations in stresstransfer would explain the local maxima and minima in the fibrilstrain measured in stiffened KASW-treated MCT observed in Fig.3A. In this regard, the recent finding of collagen molecular short-ening—due to water content changes—leading to large tensilestresses in tendon may be of relevance (59).These findings shed light on the synergistic action at the nano-

scale enabling mutability in MCT. By itself, the stress in the stiff-ened (or softened) interfibrillar matrix is by no means sufficient toaccount for the change in tissue stiffness. However, due to the largeanisotropy and surface to volume ratio of the fibrils, considerablecontact area exists for binding of the interfibrillar proteins andglycosaminoglycans to the fibrils. Such binding is likely to occur atthe gap zones (separated by D ∼ 65–67 nm) identified previously asputative binding sites for proteoglycans in collagens (60). As aresult, the total stress transferred to the collagen fibrils is effectivelyamplified by the anisotropy factor ρ ∼ 1,000. Small changes in themechanical properties of the interfibrillar matrix are thus amplifiedto apply considerable stress to the elastic fibrils, as a consequenceof which the fibrils are more effectively recruited to bear load. Priorbiochemical evidence suggests the involvement of proteins liketensilin and NSF in the stiffening of the matrix (31, 35, 37). It isprobable that these proteins physically cross-link to the existingglycosaminoglycans and proteoglycans, which in turn are bound tothe gap zones in collagen fibrils. It is conjectured that these pro-teins are acting like a bridge, binding the proteoglycan/GAG sitesof two adjacent fibrils together (12, 29). In this manner, increasedmatrix stiffening combined with effective shear load transfer to thefibrils enables MCT to undergo considerable (by a factor of ∼25;Fig. 6B) changes in tissue stiffness. The physical cross-linkingprocess is a time-dependent one (Fig. 1E), most likely due to dif-fusion of proteins secreted by the JLCs in between the collagenfibrils, combined with progressive occupation of binding sites forthese proteins in the interfibrillar matrix. This process of increasedinterfibrillar stiffening, enabling alteration of mechanical behavior,is likely to be a general property of MCT in echinoderms.

ConclusionIn summary, in this first direct measurement of the nanoscalefibrillar deformation mechanisms of MCT, we have demon-strated that the mutability of mechanical properties in thisunique invertebrate collagenous tissue is achieved solely byinterfibrillar matrix cross-linking and uncross-linking. Increasinginterfibrillar cohesion in the mechanically active state leads tonearly a 50-fold increase in fibrillar stress, underpinning thetransition of MCT from a soft to a stiff state. A greater fibrillarstress recruitment, mediated by shear transfer from the interfi-brillar matrix, leads to an over 20-fold increase in tissue modulusover the timescale of several seconds. The use of in situ X-raymethods together with mechanical testing has enabled us toquantify both the material-level mechanisms and the constitutiveproperties of the components of MCT as they undergo thesechanges in real time. The ability of MCT to undergo such largechanges in stiffness with minimal dimensional changes and solelyby increasing fibrillar recruitment highlights the potential of suchfibrillar-hydrogel composites to act as dynamic biomaterials thatcan change their mechanical state rapidly. Such materials couldfind application as new pharmacological agents, whereas thedesign of a new class of mechanically responsive nanocomposites(12) could enable energetically efficient biomaterials and devicesthat not only provide structural support but also can dynami-cally adjust their properties to the external environment forresponding to different demands. The approach opens up severalpromising further avenues of investigation; e.g., alterations in themolecular-level diffraction patterns would provide combinedmolecular as well as fibrillar real-time structural information,and the use of novel synthetic peptides mimicking the stiffeningor softening agents in MCT could be tested in situ for efficacy(61). Finally the combination of experiments with molecularmodeling methods would enable us to link mechanisms at thesmallest structural levels to macroscopic behavioral patterns.

Materials and MethodsDissection of Sea Cucumber Body Wall Preparations. Specimens of the seacucumbers (Holothuria leucospilota) were obtained from a commercialwholesaler (Marine Life) and delivered to the synchrotron SAXD beamline(in tanks of artificial sea water) a few hours before use. To prepare sea cu-cumber samples for the mechanical testing in different ionic solutions, wefollowed a protocol similar that used in previous studies of the mechanics ofsea cucumber dermis (19, 62). Specifically, after letting the sea cucumbersrest for 1 h in sea water, samples from the white central part of the bodywall dermis (Fig. 1) were prepared. The attached viscera and muscle layers onthe inside were pulled off with forceps, and the pigmented outer dermis wasremoved using razor blades, leaving only the collagenous part. Rectangular-shaped collagenous tissue pieces (10–20 mm × 1.0 mm × 1.0 mm) were cutout, using a specialized construct with twin razor blades fixed on either sideof a 1-mm-thick steel section, to keep the thickness constant to 1 mm. Thesectioning was done in the longitudinal direction of the body wall, used fortensile testing (Fig. 1 C and D). Whereas the total length of the samplevaried between specimens, the gauge length was kept constant to 6 mmduring the tensile testing (described in the next section). Samples wererinsed in ASW after sectioning. After the sections were prepared, and beforemechanical testing, all samples were allowed to relax in ASW for 1 h. Fol-lowing this, specimens were incubated in ASW (control), KASW (stiff; high[K+]), or CaF-ASW (soft) for 1.5 h. Compositions of these three solutions(ASW, KASW, and CaF-ASW) followed the protocol described previously byMotokawa and Tsuchi (14). As expected, this procedure resulted in relativeelevation and reduction of stiffness for KASW-treated and CaF-ASW–treateddermis, consistent with other studies of chemically treated MCT (30). Allstudies were carried in accordance with the Animals (Scientific Procedures)Act 1986 of the United Kingdom, including revision 2013; invertebrates(except cephalopods) are not considered protected species under the Act.

In Situ Mechanical Testing with SAXD. A compact micromechanical tester,designed by our group (41), specialized for holding biological tissues and ca-pable of being fixed on the sample stage of a synchrotron SAXD beamline,was used. The tester contains a load cell (100 N rating), with attached amplifier(RDP Electronics Ltd.). Strain is applied by displacement of a DC motor with an

8 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1609341113 Mo et al.

encoder (M-126.DG; Physik Instrumente). A customized LabVIEW (NationalInstruments) interface on a control PC was used to control the applied tissuestrain and strain rates. Adherence of the sample to the tensile tester grips wasimproved by using sandpaper of various grades between the tissue and thegrip. The machine compliance of the tester was measured using a thick steelsection. Compliance was found to be negligible compared with the stiffness ofthe sea cucumber body wall tissue under investigation. Engineering tissuestrain («T0) was calculated from the ratio of the displacement of the samplegrips to the unstressed gauge length ∼6 mm. As MCT is a soft tissue capable ofconsiderable elongation, the engineering tissue strain («T0) was converted intotrue tissue strain (eT), using eT = ln(1 + «T0) (Supporting Information, Engi-neering Tissue Strain «T0 and True Strain «T) (63). Tissue stress σ was obtainedby dividing force by sample area (1.0 mm2). Rate of increase of stress withtissue strain [tangent modulus ET (63)] was obtained from a linear regressionbetween σ and «T with a moving window of ∼Δ«T = 0.5%.

Combined microtensile deformation experiments with time-resolved acquisi-tion of SAXDpatternswere carried out at theHigh Brilliance ID02 beamline at theEuropean Synchrotron Radiation Facility (ESRF). Body wall preparations weremounted in themicrotensile tester immediately after incubation in test solutions.Samples with gauge length ∼6 mm were stretched to failure at a constant ve-locity of 0.01 mm/s (corresponding to a strain rate of ∼0.167%/s). Samples werekept hydrated by dropwise addition of the incubation solution during the test.SAXD patterns were acquired with a FReLoN CCD detector (64) with a 0.5-sexposure time, using a highly collimated synchrotron X-ray beam [beam size20 μm (height) × 25 (width) μm] at sample and detector positions, wavelengthλ = 0. 9951 Å (X-ray energy 12.46 keV) at a sample-to-detector distance of1,006.8 ± 1.0 mm determined with silver behenate at the sample position.Each SAXD pattern had a resolution of 2,048 × 2,048 pixels and a pixel areaof 23.63 × 23.97 μm2. SAXD patterns were collected continuously up tofailure of the specimen, with an interval between acquisitions of ∼1.5%strain. Each SAXD pattern is therefore acquired on a tissue location that is∼6 mm × 0.015 = 90 μm shifted from the previous measurement. The beamdiameter is much smaller (approximately one-fifth of this shift) and thereis thus no overlap of the beam onto tissue locations across SAXD mea-surements. As a result radiation damage, due to multiple exposure of thebeam to the same tissue location, is minimized. Radiation damage ofprotein assemblies and solutions occurs via a combination of free radicalsproduced by water photolysis with free radicals from the proteins, leadingto protein unfolding, aggregation, or breakage (65). Our strategy ofcontinuous sample movement to avoid repeat exposures of the same pointis one of several successful approaches to minimize radiation damage (65).Other methods, including continuously replacing the sample (e.g., con-tinuous flow in liquids) or adding radical scavengers to solutions (assumingno structural consequences) (65), are not applicable in the case of strain-stressed tissues considered here. Failure of the specimen usually occurredbetween 50% and 70% strain. As a result, typically about 40 patterns persample (60%/1.5% = 40) were acquired.

Determination of Fibril Strain from SAXD. Two-dimentional SAXD patterns ofsea cucumber body wall collagen were obtained (Fig. 2B) and averagedazimuthally (in the angular plane of the X-ray detector) to obtain theBragg peaks arising from the D periodicity of SAXD fibrils. The azimuthalaverage of the intensity provides a 1D intensity profile I(q) (q being thewavevector), which has characteristic Bragg peaks at integer multiplesof 2π/D. The software package Fit2D (66) was used, with the CAKE/INTEGRATE command, to carry out the integration. The fifth-order peakwas used for fitting as it had the strongest peak intensity among the vis-ible Bragg orders, enabling accurate peak fitting and determination ofpeak shifts [Supporting Information, Radial Intensities of Different Me-ridional Bragg Peaks in I(q) and Fig. S1]. To center the pattern around theclear fifth-order peak at ∼0.48 nm−1, inner and outer wavevector limits of0.45 nm−1 and 0.50 nm−1, respectively, were used. Subsequently, the se-lected fifth-order peak was fitted by a Gaussian function with a linearbackground term to account for the diffuse intensity scattering

IðqÞ= I05 exp�−12

�q−q05

w

�2�+ I00 + I′01q. [4]

Here I05, q05, and w represent the peak amplitude, peak position, and me-ridional peak width, respectively (I00 and I′01 are diffuse background terms).

The D period was calculated from the relation D = 5 × 2π/ q05. The per-centage changes in D value at nonzero external force (relative to the un-stressed state) provide the critical fibril strain parameter «F. The methoddescribed here has been used extensively by us for vertebrate collagenoustissues, specifically for bone and tendon (41, 42, 46, 67)

«F =Dð«T Þ−Dð«T =0Þ

Dð«T = 0Þ =q0ð«T = 0Þq0ð«T Þ − 1. [5]

To obtain the fibril strain ratio «F/«T for each sample, a linear regression of fibrilstrain «F vs. tissue strain «T was carried out (Sigma Plot; Systat Software), and theslope of the linear regression provided «F/«T (41, 44, 46, 48). The values of «F/«Tfor each sample from the different treatment groups are given in Table S1.

Determination of Fibril Orientation Measured from SAXD. In a complementarymanner to fibril strain, the angular fibril distribution was calculated from theazimuthal intensity profile I5(χ) of the fifth-order Bragg reflection of the col-lagen D spacing. The azimuthal profile was calculated by first integrating (usingthe Fit2D/CAKE command) the 2D intensity pattern radially in a narrow band ofwavevectors around the peak position q05 ∼ 0.48 nm−1 of the fifth-order re-flection, i.e., over the wavevector range 0.45–0.50 nm−1. The background-cor-rected azimuthal intensity distribution Icorrected(χ) was calculated by firstaveraging the azimuthal intensity profiles in two rings around the fifth-orderpeak position and subtracting the averaged intensity from the center (peak)ring, as described earlier for bone (68), and is shown in Fig. 2 C, 1 and Fig. S2

IcorrectedðχÞ= IoriginalðχÞ− 12ðIouterðχÞ+ IinnerðχÞÞ. [6]

To ensure the full azimuthal width of the peaks was captured, the intensityprofile was calculated over the full circle (0°–360°). Angular coordinatescorresponding to high values of Icorrected(χ) denote a greater proportion offibrils along the specified azimuthal angle. The profile Icorrected(χ) was thenfitted to a function with two Gaussian peak profiles, separated by 180°

IcorrectedðχÞ= I0χ

�exp

�−12

�χ − χ0Δχ0

�2�+ exp

�−12

�χ − χ0 − π

Δχ0

�2��. [7]

The parameter χ0 defines the main direction of orientation of the fibrils,whereas Δχ0 is a parameter characterizing the width of the distribution andI0χ is an amplitude term proportional to the total SAXD intensity of the fifth-order reflection. It is noted that whereas applied forces will induce shifts inq05 over the course of the test, these will turn out (shown in Results) to besufficiently small such that the same narrow band around the initial peakposition can be used over the entire test.

A collagen fibril distribution with a narrow angular width (correspondingto well-oriented fibrils) is characterized by a low value Δχ0 and twin sharppeaks in Icorrected(χ) above a low baseline intensity, whereas a distributionwith a wide angular dispersion in fibril orientation is characterized by highΔχ0 and a nearly constant Icorrected(χ). We define a dimensionless parameterυ, derived from Icorrected(χ), which can be used to determine whether thefibril distribution is narrow or broad

υ=  SDðIcorrectedðχÞÞ=MeanðIcorrectedðχÞÞ. [8]

It can be seen that an isotropic (wide) fibril angular distribution, corre-sponding to a nearly constant Icorrected(χ), will have υ ∼ 0, whereas υ willincrease as the angular width reduces.

ACKNOWLEDGMENTS. We thank Jun Ma (School of Engineering and MaterialsScience, University of London) for help designing the software interface for thetensile tester, Vince Ford (School of Engineering andMaterials Science, Universityof London) for technical support in manufacturing parts for the tensile tester,and T. Narayan (Beamline ID02, ESRF) for excellent support during the beamtime.The principal SAXD experiments were performed on beamline ID02 at theESRF. J.M. is supported by the China Scholarship Council. H.S.G. and M.R.E.acknowledge support from the Engineering and Physical Sciences ResearchCouncil (EP/J501360/1), the Biotechnology and Biological Sciences ResearchCouncil (BB/M001644/1), and the Royal Society through the Equipment Grantscheme (SEMF1A6R). L.M.B., H.S.G., and M.R.E. acknowledge support from theInstitute of Bioengineering at Queen Mary University of London.

1. Egan P, Sinko R, LeDuc PR, Keten S (2015) The role of mechanics in biological and bio-

inspired systems. Nat Commun 6:7418.2. Meyers MA, Chen PY, Lin AYM, Seki Y (2008) Biological materials: Structure and

mechanical properties. Prog Mater Sci 53(1):1–206.

3. Fung Y-c (2013) Biomechanics: Mechanical Properties of Living Tissues (Springer Sci-

ence & Business Media, Berlin).4. Weinkamer RDJ, Brechet Y, Fratzl P (2013) All but diamonds—Biological materials are

not forever. Acta Mater 61(3):880–889.

Mo et al. PNAS Early Edition | 9 of 10

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

ENGINEE

RING

PNASPL

US

5. Eyre DR (1980) Collagen: Molecular diversity in the body’s protein scaffold. Science207(4437):1315–1322.

6. Harrington MJ, Masic A, Holten-Andersen N, Waite JH, Fratzl P (2010) Iron-clad fibers:A metal-based biological strategy for hard flexible coatings. Science 328(5975):216–220.

7. Fratzl P, Fratzl-Zelman N, Klaushofer K (1993) Collagen packing and mineralization.An x-ray scattering investigation of turkey leg tendon. Biophys J 64(1):260–266.

8. Ciarletta P, Ben Amar M (2009) A finite dissipative theory of temporary interfibrillarbridges in the extracellular matrix of ligaments and tendons. J R Soc Interface 6(39):909–924.

9. Julicher F, Ajdari A, Prost J (1997) Modeling molecular motors. Rev Mod Phys 69(4):1269–1281.

10. Motokawa T (1984) Connective-tissue catch in echinoderms. Biol Rev Camb Philos Soc59(2):255–270.

11. Wilkie IC (1984) Variable tensility in echinoderm collagenous tissues - a review. MarBehav Physiol 11(1):1–34.

12. Wilkie I (2005) Mutable collagenous tissue: Overview and biotechnological perspective.Echinodermata, ed Matranga V (Springer-Verlag, Berlin), pp 221–250.

13. Motokawa T (1994) Effects of ionic environment on viscosity of Triton-extracted catchconnective tissue of a sea cucumber body wall. Comp Biochem Physiol B 109(4):613–622.

14. Motokawa T, Tsuchi A (2003) Dynamic mechanical properties of body-wall dermis invarious mechanical states and their implications for the behavior of sea cucumbers.Biol Bull 205(3):261–275.

15. Ribeiro AR, et al. (2011) New insights into mutable collagenous tissue: Correlationsbetween the microstructure and mechanical state of a sea-urchin ligament. PLoS One6(9):e24822.

16. Benedetto CD, et al. (2014) Production, characterization and biocompatibility ofmarine collagen matrices from an alternative and sustainable source: The sea urchinParacentrotus lividus. Mar Drugs 12(9):4912–4933.

17. Wilkie IC, et al. (2015) Mechanical properties of the compass depressors of the sea-urchin Paracentrotus lividus (Echinodermata, Echinoidea) and the effects of enzymes,neurotransmitters and synthetic tensilin-like protein. PLoS One 10(3):e0120339.

18. Birenheide R, Motokawa T (1996) Contractile connective tissue in crinoids. Biol Bull191(1):1–4.

19. Takemae N, Nakaya F, Motokawa T (2009) Low oxygen consumption and high bodycontent of catch connective tissue contribute to low metabolic rate of sea cucumbers.Biol Bull 216(1):45–54.

20. Barbaglio A, et al. (2012) The mechanically adaptive connective tissue of echino-derms: Its potential for bio-innovation in applied technology and ecology. MarEnviron Res 76:108–113.

21. Jordan H (1914) Die holothurien als hohlorganartige tiere und die tonusfunktionihrer muskulatur [The holothurians as hollow organ-like animals and their muscletone]. Zool Jahrb Abt 34:365–436.

22. Motokawa T (1981) The stiffness change of the holothurian dermis caused bychemical and electrical stimulation. Comp Biochem Physiol C 70(1):41–48.

23. Motokawa T (1982) Factors regulating the mechanical properties of holothuriandermis. J Exp Biol 99:29–41.

24. Wilkie IC, Carnevali MDC, Bonasoro F (1992) The compass depressors of Para-centrotus-Lividus (Echinodermata, Echinoida) - ultrastructural and mechanical aspectsof their variable tensility and contractility. Zoomorphology 112(3):143–153.

25. Hidaka M, Takahashi K (1983) Fine structure and mechanical properties of the catchapparatus of the sea-urchin spine, a collagenous connective tissue with muscle-likeholding capacity. J Exp Biol 103(1):1–14.

26. Diab M, Gilly WF (1984) Mechanical properties and control of non-muscular catch inspine ligaments of the sea urchin, Strongelocentrotus franciscanus. J Exp Biol 111(1):155–170.

27. Szulgit G, Shadwick R (1994) The effects of calcium chelation and cell perforation onthe mechanical properties of sea urchin ligaments. Echinoderms Through Time, Pro-ceedings of the Eighth International Echinoderm Conference, eds David B, Guille A,Féral J-P, Roux M (CRC Press, Boca Raton, FL), pp 887–892.

28. Trotter JA, Koob TJ (1989) Collagen and proteoglycan in a sea urchin ligament withmutable mechanical properties. Cell Tissue Res 258(3):527–539.

29. Trotter JA, Lyons-Levy G, Thurmond FA, Koob TJ (1995) Covalent composition ofcollagen fibrils from the dermis of the sea cucumber, Cucumaria frondosa, a tissuewith mutable mechanical properties. Comp Biochem Physiol A 112(3–4):463–478.

30. Motokawa T (2011) Mechanical mutability in connective tissue of starfish body wall.Biol Bull 221(3):280–289.

31. Yamada A, Tamori M, Iketani T, Oiwa K, Motokawa T (2010) A novel stiffening factorinducing the stiffest state of holothurian catch connective tissue. J Exp Biol 213(Pt 20):3416–3422.

32. Oji T, Okamoto T (1994) Arm autotomy and arm branching pattern as anti-predatoryadaptations in stalked and stalkless crinoids. Paleobiology 20(1):27–39.

33. Trotter JA, Chapman JA, Kadler KE, Holmes DF (1998) Growth of sea cucumber col-lagen fibrils occurs at the tips and centers in a coordinated manner. J Mol Biol 284(5):1417–1424.

34. Trotter JA, et al. (1996) Stiparin: A glycoprotein from sea cucumber dermis that ag-gregates collagen fibrils. Matrix Biol 15(2):99–110.

35. Koob TJ, Koob-Emunds MM, Trotter JA (1999) Cell-derived stiffening and plasticizingfactors in sea cucumber (Cucumaria frondosa) dermis. J Exp Biol 202(Pt 17):2291–2301.

36. Cluzel C, Lethias C, Humbert F, Garrone R, Exposito JY (2001) Characterization of fi-brosurfin, an interfibrillar component of sea urchin catch connective tissues. J BiolChem 276(21):18108–18114.

37. Tipper JP, Lyons-Levy G, Atkinson MA, Trotter JA (2002) Purification, characterizationand cloning of tensilin, the collagen-fibril binding and tissue-stiffening factor fromCucumaria frondosa dermis. Matrix Biol 21(8):625–635.

38. Trotter JA, et al. (1999) Collagen fibril aggregation-inhibitor from sea cucumberdermis. Matrix Biol 18(6):569–578.

39. Ribeiro AR, et al. (2012) Matrix metalloproteinases in a sea urchin ligament withadaptable mechanical properties. PLoS One 7(11):e49016.

40. Cluzel C, Lethias C, Garrone R, Exposito JY (2004) Distinct maturations of N-propep-tide domains in fibrillar procollagen molecules involved in the formation of hetero-typic fibrils in adult sea urchin collagenous tissues. J Biol Chem 279(11):9811–9817.

41. Gupta HS, et al. (2013) Intrafibrillar plasticity through mineral/collagen sliding is thedominant mechanism for the extreme toughness of antler bone. J Mech BehavBiomed Mater 28:366–382.

42. Screen H, Seto J, Krauss S, Boesecke P, Gupta H (2011) Extrafibrillar diffusion andintrafibrillar swelling at the nanoscale are associated with stress relaxation in the softcollagenous matrix tissue of tendons. Soft Matter 7(23):11243–11251.

43. Zimmermann EA, et al. (2013) Mechanical adaptability of the Bouligand-type struc-ture in natural dermal armour. Nat Commun 4:2634.

44. Zimmermann EA, et al. (2011) Age-related changes in the plasticity and toughness ofhuman cortical bone at multiple length scales. Proc Natl Acad Sci USA 108(35):14416–14421.

45. Hulmes DJ, Jesior J-C, Miller A, Berthet-Colominas C, Wolff C (1981) Electron mi-croscopy shows periodic structure in collagen fibril cross sections. Proc Natl Acad SciUSA 78(6):3567–3571.

46. Gupta HS, et al. (2006) Cooperative deformation of mineral and collagen in bone atthe nanoscale. Proc Natl Acad Sci USA 103(47):17741–17746.

47. Stock SR, Almer JD (2009) Strains in bone and tooth via high energy X-ray scattering.Bone 44(2):S270.

48. Puxkandl R, et al. (2002) Viscoelastic properties of collagen: Synchrotron radiationinvestigations and structural model. Philos Trans R Soc Lond B Biol Sci 357(1418):191–197.

49. Gasser TC, et al. (2012) Spatial orientation of collagen fibers in the abdominal aorticaneurysm’s wall and its relation to wall mechanics. Acta Biomater 8(8):3091–3103.

50. Linari M, et al. (2015) Force generation by skeletal muscle is controlled by mecha-nosensing in myosin filaments. Nature 528(7581):276–279.

51. Yang W, et al. (2015) On the tear resistance of skin. Nat Commun 6:6649.52. Capadona JR, Shanmuganathan K, Tyler DJ, Rowan SJ, Weder C (2008) Stimuli-

responsive polymer nanocomposites inspired by the sea cucumber dermis. Science319(5868):1370–1374.

53. Eppell SJ, Smith BN, Kahn H, Ballarini R (2006) Nano measurements with micro-devices: Mechanical properties of hydrated collagen fibrils. J R Soc Interface 3(6):117–121.

54. Fratzl P (2008) Collagen: Structure and Mechanics, an Introduction (Springer,Amsterdam).

55. Jäger I, Fratzl P (2000) Mineralized collagen fibrils: A mechanical model with astaggered arrangement of mineral particles. Biophys J 79(4):1737–1746.

56. Gao H, Ji B, Jager IL, Arzt E, Fratzl P (2003) Materials become insensitive to flaws atnanoscale: Lessons from nature. Proc Natl Acad Sci USA 100(10):5597–5600.

57. Fratzl P (2007) Nature’s hierarchical materials. Prog Mat Sci 52(8):1263–1334.58. Trotter JA, Thurmond FA, Koob TJ (1994) Molecular structure and functional mor-

phology of echinoderm collagen fibrils. Cell Tissue Res 275(3):451–458.59. Masic A, et al. (2015) Osmotic pressure induced tensile forces in tendon collagen. Nat

Commun 6:5942.60. Scott JE (1990) Proteoglycan:collagen interactions and subfibrillar structure in colla-

gen fibrils. Implications in the development and ageing of connective tissues. J Anat169:23–35.

61. Elphick MR (2012) The protein precursors of peptides that affect the mechanics ofconnective tissue and/or muscle in the echinoderm Apostichopus japonicus. PLoS One7(8):e44492.

62. Tamori M, et al. (2006) Tensilin-like stiffening protein from Holothuria leucospilotadoes not induce the stiffest state of catch connective tissue. J Exp Biol 209(Pt 9):1594–1602.

63. Vincent JF (2012) Structural Biomaterials (Princeton Univ Press, Princeton, NJ).64. Narayanan T, Diat O, Boesecke P (2001) SAXS and USAXS on the high brilliance

beamline at the ESRF. Nucl Instrum Methods Phys Res A 467:1005–1009.65. Jeffries CM, Graewert MA, Svergun DI, Blanchet CE (2015) Limiting radiation damage

for high-brilliance biological solution scattering: Practical experience at the EMBL P12beamline PETRAIII. J Synchrotron Radiat 22(2):273–279.

66. Hammersley AP (2016) FIT2D: A multi-purpose data reduction, analysis and visuali-zation program. J Appl Crystallogr 49:646–652.

67. Karunaratne A, et al. (2012) Significant deterioration in nanomechanical quality oc-curs through incomplete extrafibrillar mineralization in rachitic bone: Evidence fromin-situ synchrotron X-ray scattering and backscattered electron imaging. J BoneMinerRes 27(4):876–890.

68. Karunaratne A, et al. (2016) Multiscale alterations in bone matrix quality increasedfragility in steroid induced osteoporosis. Bone 84:15–24.

10 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1609341113 Mo et al.